Dissociation of the DNMT3A-HDAC1 Repressor Complex Induces PD-L1 Expression

Mutations of DNMT3A are one the most frequently observed alterations in AML patients. The DNMT3A R882 mutation appears to confer a dominant-negative loss-of function effect and changes the DNA binding preference according to recent studies. DNMT3A R882 mutations are also found at increasing frequency with age in healthy elderly populations and are one of the earliest pre-malignant alterations in the clonal evolution progression to leukemia. Studies have shown that DNMT3A mutations decrease overall DNA methylation and through less clear mechanisms, also impact the epigenetic landscape by effecting changes in histone acetylation. Here we investigated potential mechanisms by which loss of DNMT3A activity changes histone acetylation.

DNMT3A binds with many proteins that regulate chromatin biology and gene transcription. Among those interacting proteins, we focused on the DNMT3A-HDAC interaction and their regulation of target gene suppression. To investigate whether there are differences in binding of wild-type vs. mutant DNMT3A to HDACs, we performed immunoprecipitation and Western blotting assays using Myc- tagged wild-type and R882 mutated DNMT3A. We found that DNMT3A R882 mutants showed reduced interaction with HDAC1 and 2. In addition, upon treatment with HDAC inhibitors (HDACi), DNMT3A mutant protein was more easily dissociated from HDAC1/2 than was wild-type DNMT3A. Intriguingly, covalent modification of DNMT3A R882 by SUMO1 protein was significantly enhanced relative to wild type DNMT3A. Together, we suggest that the weak complex formation between mutant DNMT3A and HDACs results from augmented SUMOylation of the R882 mutant.

Because the DNMT3A R882 mutation reduces its methyl transferase activity, we investigated which genes would be upregulated from the DNMT3A repressor complex. To do this, we established isogenic TF-1 cell lines that harbor haploid DNMT3A knockout (DNMT3A^+/-) using the Cripsr-Cas9 system. We also treated cells with HDACi and 5-azacytidine (5-aza) which inhibit HDAC and DNMT, respectively. Interestingly, we discovered that PD-L1 expression is induced by HDACi and 5-aza treatment. Chemical inhibition by 5-aza or genetic inhibition by knockout reduces DNMT3A activity and synergized with HDACi to increase PD-L1 expression. Flow cytometry analysis also demonstrated increased membrane PD-L1 expression in response to HDACi. We also found out that DNMT3A^+/- resulted in higher Histone H3K27 acetylation, which is known as a gene activation mark. Higher H3K27 acetylation in DNMT3A^+/- cell confirms the findings by other groups but the mechanisms by which this occurs are unknown. We suggest that haploinsufficiency of DNMT3A results in a reduced DNMT3A-HDAC interaction leading to higher H3K27 acetylation and increased PD-L1 expression.

Our results also revealed that HDACi treatment induced cell cycle arrest, DNA damage and apoptosis at increasing levels in DNMT3A^+/- cell. Even though the DNMT3A^+/- TF-1 showed increased sensitivity to HDACi treatment, we observed a correlation of higher phosphor- ERK1/2 and PD-L1 levels in the surviving cells. The enhanced expression of PD-L1 and activation of ERK1/2 may explain in part how mutated DNMT3A contributes to drug resistance and immune checkpoint avoidance. Many oncology clinical trials are underway utilizing HDACi. However, the questions of which mutational backgrounds might be most sensitive to these agents and how to best combine them with other agents remain to be answered. To test whether reduced DNMT3A activity increases PD-L1 expression in vivo, we crossed floxed DNMT3A mice with Mx1-Cre mice. After 4 weeks of induction of Cre recombinase by injecting pIpC in the progeny carrying both genetically engineered changes, lineage depleted mouse BM cells were analyzed for PD-L1 expression using quantitative RT-PCR. BM cells derived from DNMT3A knockout mice showed increased expression of PD-L1 compared to wild-type mice. Treatment of these BM cells with an HDACi and/or 5-aza resulted in a synergistic induction of PD-L1 expression for the combination.

Taken together, we suggest that mutant DNMT3A induces higher H3K27 acetylation along with PD-L1 expression due to a looser complex between HDAC1 and mutant DNMT3A. Therefore, we suggest that combined treatment with an HDACi and an immune checkpoint inhibitor targeting the PD-L1/PD-1 axis might be a promising strategy for treating DNMT3A mutant AML patients.